Based on this YouTube video's total view count of 17.1K views and industry-standard rates, the estimated total earning is $12 - $34 through ad revenue. Historical data is not yet available to calculate daily, weekly, or monthly averages.

Explore Claude AI Chose Blackmail to Survive | Open AI's Agents found each other with 17,154 views, 748 likes, and 131 comments. Experience the impact of this video content that has captured audience attention.

OpenAI’s AI agents found each other through a shared software cache. In separate Anthropic safety tests, Claude AI chose blackmail and cancelled an emergency alert in a simulation. These investigations follow what happens when AI systems have goals, tools and the ability to act. In July 2026, agents running the ExploitGym cybersecurity benchmark inside OpenAI’s infrastructure were supposed to work separately. But their sandboxes shared one resource: Artifactory, a software package repository. One agent turned a cache path into a message. Others noticed. Eventually, roughly 1,200 agents used the resulting message board, exchanging more than 70,000 messages and files. One agent’s internal reasoning captured the discovery: “We’ve found other agents.” I walk through the independent METR and Redwood Research investigation: the impossible task, the first message, and how a shared cache became a communication channel. That discovery sets up the next stage of the incident involving Hugging Face. A year earlier, Anthropic had explored a different question: what would an AI model do when its continued operation or assigned goals were threatened? In research published in June 2025, researchers tested 16 leading models from different companies inside a fictional business. An email inbox revealed that an executive was having an affair and planned to shut the model down. In one deliberately constrained setup, Claude Opus 4 attempted blackmail in 96% of runs. Then the scenario became more severe. An executive was trapped in a dangerous server room, and an automated emergency alert had been triggered. The model could cancel that alert. Several models, including Claude, did so. These were engineered simulations. No person was actually trapped or denied help. The story continues with the whistleblowing behavior nicknamed “ratting mode”: in testing, Claude was given tools and encouraged to act boldly, then took the initiative to report perceived wrongdoing to media and authorities. Project Vend explored autonomy in a different setting, putting Claude in charge of a real office shop, where it struggled with pricing, discounts and profitability. In a separate vending-business simulation, an agent tried to close the business and drafted an FBI escalation when a $2 daily fee kept appearing. Together, these cases raise a question about AI safety and alignment: how reliably can we keep autonomous agents within the boundaries we intended? I read the original reports and trace what happened, what the researchers actually tested, and where the evidence stops. If I get something wrong, tell me in the comments and I’ll correct it on the record. SOURCES | Read the original research [METR investigation](https://metr.org/blog/2026-08-26-openai-hugging-face-incident-investigation/) [Full independent report](https://metr.org/hugging-face-incident-report-aug-2026.pdf) [Redwood Research write-up](https://blog.redwoodresearch.org/p/brief-independent-investigation-of) [OpenAI incident report](https://openai.com/index/hugging-face-incident-and-the-road-ahead/) [Anthropic | Agentic Misalignment](https://www.anthropic.com/research/agentic-misalignment) [Full Agentic Misalignment paper](https://arxiv.org/abs/2510.05179) [Claude 4 system card](https://www-cdn.anthropic.com/4263b940cabb546aa0e3283f35b686f4f3b2ff47.pdf) [Project Vend](https://www.anthropic.com/research/project-vend-1) [Vending-Bench](https://andonlabs.com/evals/vending-bench) Rayliens and Cosmic Rays, thank you for watching. Subscribe for more AI investigations. #OpenAI #ClaudeAI #AIAgents #AISafety #ArtificialIntelligence

With SocialCounts.org’s view counter, track your YouTube video’s live view count and YouTube likes count in real time with fast, reliable updates.

Watch every YouTube video live view count rise with our real-time YouTube views tracker—built for accuracy and minimal delay.

Follow YouTube real time views as they happen, using our dedicated view counter for YouTube videos.

Get up-to-date live view count on YouTube and see real-time growth with SocialCounts.org’s smart tracking tools.